Vector Database
A database designed to store and search embeddings — the numerical representations of meaning used by AI models.
A vector database stores embeddings and lets you search them by similarity rather than exact match. Instead of asking "find the row where name = X", you ask "find the 10 embeddings most similar to this one." This makes it possible to build semantic search, recommendation engines, and RAG systems.
A regular database is like a filing cabinet — everything has a label and you find things by looking up the exact label. A vector database is like a library organised by topic similarity — you walk in with a book and the librarian finds you the 10 most thematically similar books, even if none of them share a title or author.
Vector databases aren't a replacement for traditional databases. They're a complement. Most production AI systems use both — a vector database for semantic retrieval and a relational database for structured data like user accounts, transactions, and metadata.